Power-System Modelling for Computation

Page status: overview of the teaching draft and complete reference library.

From circuit equations to optimization and software

Two parallel lines each have a 100 A rating. Replacing them by one line with the same total admittance and a 200 A rating can admit an operating point that overloads one of the original lines. The calculation looks consistent until the member currents are recovered.

Start with A plausible model gives the wrong answer. Derive the failure, run the calculation, and repair the rating. The example leads to the book's central question: which physical quantities, constraints, and decisions survive the model you compute with?

Read the book

The teaching route develops eight skills: diagnose a model failure; construct equipment equations; handle conductors and grounding; choose a computational graph; transform and recover a model; preserve constraints and decisions; assess numerical evidence; and defend an end-to-end study.

Follow How to use this book, or download the core PDF. Read the author's preface for the experience that motivated the work. The route is a selective draft: some parts still use longer reference chapters while their teaching versions are developed.

The intended readers are scientifically demanding power engineers and computer scientists or operations researchers developing power-system applications. Bring linear algebra, complex arithmetic, and elementary circuit laws. The first experiment requires only Python 3; Julia is used in the larger studies.

Run and inspect

The computational case guide connects lessons to commands, expected evidence, and questions to investigate. The study workbook provides a structure for recording and defending a complete modelling experiment.

Consult the reference library

The HTML library retains the full specialist treatment and evidence records. Use the knowledge-base index, terminology, and chapter status to locate a claim and its scope. Scope and thesis explains the proposed representation architecture and the limits of current application coverage.

The current draft has no claims recorded as externally reviewed. Derivations, repository tests, independent numerical implementations, and human review are distinct evidence categories. The evidence map reports the maintained inventory.

For optional language-model access to the same sources, see the ChatGPT guide or Claude guide.